EPTLet T = number of trolls in a game
It's tough to estimate how many games belong in the former category and how many are in the latter. I suspect that for each n in 0->10, P(T=n) varies with regard to skill range, region, date, time, etc, quite a bit. I also would not assume that one player being a troll in a game is an independent event of other players being a troll in the game. Perhaps one day I'll scan dotabuff to see if I can estimate these correlations, but today I'm mostly focusing on how this troll factor affects each game. We'll be looking at games where T>0.
I suspect the vast majority of these games fall into T=1, but that the probability of a second troll increases dramatically once the first obvious game ruining comes out. Sometimes, a troll will be tricked into game ruining by an incorrect signal, such as a crystal maiden that is afk and gets caught out, but it otherwise still "in the game".
Reasons to troll a game could be because a number of factors that are likely not independent:
(a) you perceive it as a loss already and want it to end sooner
(b) you have a grudge against a teammate and wish to make them lose
(c) outside factors, such as internet problems causing you to d/c
(d) other factors
For cases where T>1, I hypothesize that (a) and (b) increases dramatically, but the chance of a d/c in a rather short time between the beginning of the troll to the end of the game decreases. In any case, it only takes one troll to ruin a game.
if you have a lower than average propensity for game ruining at your skill level, or if you have the additional skill of mitigating strife between teammates or focusing them on playing to win a game, even if it feels out of their control, then your perceived P(T>0) will be lower, regardless of the bracket you are in. If you tend to troll more than average, then P(T>0) will be higher, and P(T>1) will also be higher.
Additionally, by not trolling, you have a higher chance of there not being a troll on your team, where there might be one lurking on their team, giving you a better chance at winning, assuming there is at least one game ruining player in the game.
I guess the point here is that the best way to get rid of having to play with trolls in the game is simply to not be one. Another way is to address factors that contribute to trolls.
Anecdotally, I've found that it's more difficult to reduce trolling in partied matches rather than solo queue. I believe it's contributed by the fact that the friends I play with are more likely to troll than average, and also that it's less effective to use the same troll mitigation tactics more than once on the same player, so you have to be more creative.
+ Show Spoiler +
edit: 2^8 posts on LD. Am I internet famous yet?
